In the past two decades, researchers and clinicians have increasingly focused on the early characterization and identification of young children with delayed language acquisition, termed late talkers (LTs). LTs are identified on the basis of their restricted expressive language abilities when they are approximately 2 years of age; their main characteristic is a considerable delay in linguistic production with respect to other children. The criteria used to identify a child with speech delay vary from study to study and depends mainly on the instrument used to measure linguistic development. Rescorla (1989), for example, using the Language Development Survey Parent Checklist, identified as LT children who had a vocabulary of less than 50 words (or no multiword combinations) at the age of 24 months, whereas Stoel-Gammon (1989) considered as LT two children who, at the age of 21 months, produced fewer than 10 words during a 2-hr observation session. However, Thal and Tobias (1994) referred to percentiles according to the MacArthur Communicative Development Inventory (CDI), a questionnaire composed of a list of words that quantifies the level of linguistic comprehension and production.
The interest in LTs is due to the fact that they are considered to be at risk of subsequent specific linguistic problems (Bishop & Edmundson, 1987; Rescorla & Schwartz, 1990). Furthermore, it has been demonstrated that linguistic delay could conceal complex clinical problems and also represent a risk factor for the children's cognitive and psychoaffective development (Rutter & Lord, 1987). These findings led to a number of studies on children showing this phenomenon (Rescorla & Ratner, 1996; Rescorla & Schwartz, 1990; Stoel-Gammon, 1989; Whitehurst et al., 1991), which have provided a wide database on the LT population.
Several studies have attempted to identify the risk factors inherent in linguistic delay as well as the diagnostic and prognostic indexes. Significant prognostic indexes were provided by the following aspects of language development: prelinguistic babbling characteristics (Orsolini, 2000; Paul, 1991; Paul, Laszlo, McFarland, & Midford, 1993), vocabulary size from 24 to 38 months of age (Fischel, Whitehurst, Caulfield, & Debaryshe, 1989), magnitude of linguistic delay at the first consultation (Rescorla & Schwartz, 1990), lexical comprehension and symbolic gestures from 18 to 28 months of age (Thal, 1991), and use of communicative gestures (Thal & Tobias, 1994).
In a recent longitudinal study conducted on 32 Italian children with expressive or receptive–expressive linguistic delay between 18 and 33 months of age, Cipriani, Chilosi, Pfanner, Villani, and Bottari (2002) verified the predictive value of linguistic indexes that could be used to distinguish children with successful language development from children at risk of specific linguistic impairment. The results of this study indicated that the level of verbal comprehension was the significant predictive factor for subsequent language development (tested between 44 and 48 months of age). Other precocious indexes are found in vocabulary composition, and in particular, the use of verbal predicates, adjectives, and function words that are essential for sentence production. Cipriani et al. (2002) claimed that language development can be defined as atypical, with a strong risk for the subsequent outcome, if the child does not achieve the above competencies within 30 months of age.
Several studies, most of which were done with English-speaking children, as well as investigating the predictive value of linguistic indexes on subsequent linguistic development, also examined the outcome of early linguistic delay in preschool and school-aged children. According to Rescorla (2002), LTs are a subset of children with specific language impairment (SLI), but even if both groups of children have weak language systems in the presence of normal nonverbal abilities, LTs are less severely impaired and generally manifest a better outcome (Paul, 1996; Rescorla & Lee, 2000). Although only a small proportion of LTs evolve with a clinically significant linguistic delay as is in SLI, the literature also reports that LTs often remain behind their typically developing peers in certain measures of language competence, throughout middle childhood. A series of contributions show that, despite the fact that LTs catch up on norm-referenced measures of expressive language development, they continue to have difficulties with higher level linguistic tasks that could make them liable to learning and academic difficulties during the early school years (Kelly, 1998; Whitehurst et al., 1991).
In a recent study, Rescorla (2002) suggests that slow early language development reflects a predisposition for slower acquisition of a wide range of language-related skills into middle childhood. LTs did not score lower on reading tasks than comparison peers at ages 6 and 7, that is, in the early stages of reading, but they did have significantly poorer reading/spelling skills at ages 8 and 9 than their comparison peers. Additional insights into LT performance on higher level language tasks was provided by Rescorla and her collaborators (Rescorla, 1993; Rescorla, Hadicke-Wiley, & Escarce, 1993): at age 6 participants scored lower than comparison peers on verbal reasoning tasks, auditory short-term memory (STM), and word definition tasks. Sentence formulation tasks were especially challenging at 7 and 8 years of age for those children with histories of slow expressive language development. Between-group differences were also observed at 8 years of age on tasks measuring auditory processing, fluency in word retrieval, and verbal memory. Overall, results identify a number of areas of vulnerability in LTs that may potentially influence their academic achievements as they progress through the early school years.
In contrast, some studies suggest that the majority of children with a history of slow expressive language development generally recover the expressive linguistic disadvantage and do not exhibit language problems in their school years (Paul, Murray, Clancy & Andrews, 1997; Rescorla, Roberts, & Dahlsgaard, 1997; Rescorla & Schwartz, 1990). Paul et al. (1997), for example, reported that children with a history of slow expressive language development performed within normal range on the second grade speech, reading, and metaphonology measures. Rescorla (1993) reported that only 12% of LTs scored 1.5 SD below the mean on an expressive grammatical ability test at 5 years of age, whereas they scored within the normal range on the expressive language, visual motor skills, verbal, and nonverbal reasoning measures. At 6 years of age, the same children achieved age-appropriate scores on the grammatical ability tests.
Several follow-ups indicate that it is important to continue to investigate cognitive and language development in a population of children with an early history of slow expressive language development. Whereas these children consistently show some difficulty at follow-up, the nature of the difficulty, age of manifestation, and possible causes are still a matter of debate, and results are not always consistent across studies. Finally, the vast majority of these studies are based on the English language. Therefore, there is a possibility that a comparison with other languages might prove illuminating, as it has been demonstrated that a number of linguistic-related difficulties are language sensitive. With respect to Italian, to our knowledge there has only been one Italian LT follow-up study (Cipriani et al., 2002).
The aim of the present study is to extend our knowledge of language development in LTs who appear to have recovered the initial delay, in so far as they have not been diagnosed with any SLI and their parents have not reported any particular language problems. Specifically, we investigated whether these children still show mild difficulties in cognitive and linguistic capacities in the preschool years. Even if this “weakness” does not interfere with the everyday communicative use of language, it might have a negative affect on academic potential.
We evaluated some linguistic capacities (syntax and vocabulary development, phonological skills) and other aspects such as memory, attention, and impulsiveness in 4- to 6-year-olds identified as LTs at 24 months of age. Moreover, the study will provide information on a non-English language by developing our knowledge of Italian LTs.
METHOD
Participants
The participants were selected from the two samples of 26 children recruited for two studies on early predictors of language development (Fasolo & D'Odorico, 2002a, 2002b; Assanelli, D'Odorico, Franco, & Salerni, 2005). These children had been identified as LTs at 24 months of age on the basis of their mothers' responses to the Il Primo Vocabolario del Bambino Questionnaire, the Italian version of the MacArthur CDI (Caselli & Casadio, 1995); they had a vocabulary size of less than 50 words (below the 10th percentile as defined by Italian norms; Caselli & Casadio, 1995). Consent for this follow-up was obtained for 16 of these children1
Ten children were not included in the sample: one child was diagnosed with specific language impairment, the parents of four LTs declined to participate in the follow-up, and we were not able to trace the remaining five families.
For a part of the sample it has been possible to assess psychomotor development. In particular, nine LT and five AT were evaluated at 24 months of age by the Brunet–Lézine Early Scale of Psychomotor Development (Brunet & Lézine, 1967). The developmental score refers to three of the four subscales of which the test is composed: fine motor skills, posture and gross motor adaptation, and social skills.2
The fourth subscale regarding the linguistic area was not considered, because this competence was already evaluated by the Il Primo Vocabolario del Bambino Questionnaire.
None of the children suffered from neurological or psychological complications or birth problems. The children received the follow-up test at different age levels. More specifically, eight LTs and seven ATs at 4 years, four LTs and five ATs at 5 years, and four LTs and four ATs at 6 years. At the time of the follow-up none had been diagnosed as suffering from SLIs, and no particular language development difficulties had been reported by parents. The characteristics of the sample are reported in Table 1.
Materials
Verbal STM was assessed by means of two different tasks.
Repetition of nonwords
This task was taken from a battery evaluating prerequisites for reading and writing difficulties created by Cornoldi, Miato, Molin, and Poli (1995). Nonwords were created using phonotactic relations that, although possible, do not exist in the Italian language. The children were asked to repeat 25 nonwords: 5 of one syllable, 5 of two syllables, and five phonemes, 5 of two syllables and six phonemes, 5 of three syllables, and 5 of four syllables (see Appendix A). The score corresponds to the number of correct syllables repeated by each child; therefore, the maximum possible score is 60. Because two of the younger children refused to complete the tasks, we adjusted the score as the percentage of correct syllables out of the total of presented syllables.
Word span
This is a bisyllabic Word Span Task specifically devised for this study.3
We did not use a digit span task because it does not allow us to consider the load of the phonological loop device (some digits are composed of one syllable and others of two syllables).
The difference between the two tasks is based on the assumption that the nonword repetition task is not only a STM task but is mainly mediated by phonological awareness (Metsala, 1999), whereas the word span should provide a relatively pure measure of short-term verbal memory.
Phonological awareness was specifically investigated by means of three subtests of the PRCR-2 learning prerequisites battery (Cornoldi & Gruppo, 1992). Only the 5- and 6-year-olds performed these tasks, because data from a pilot study showed that they were too difficult for 4-year-olds. The tasks include the following:
- Blending phonemes: The examiner enunciated the individual phonemes of a word, and then asked the children to produce the entire word; the task was composed of 14 words (from 4 to 10 phonemes), and one point was given for each blend of phonemes. The score ranged from 0 to 74.
- Blending syllables: The examiner enunciated the individual syllables of a word and then asked the children to produce the entire word; the task was composed of 10 words of three syllables each. Two points were given for each entire word and one point for blending at least two syllables. The score ranged from 0 to 20.
- Segmentation: The children were presented with three different tasks: in the first, they had to identify and repeat the first phoneme of 10 words, in the second, they had to identify and repeat the last phoneme of 10 words, and in the third, they had to spell nine words for a total of 36 phonemes. The maximum score was 56 points.
Syntactic competence was evaluated by means of the Test di Comprensione Grammaticale per Bambini (TCGB; Chilosi & Cipriani, 1995), which is composed of 76 cards with four pictures each; the examiner read out a sentence, after which the child was asked to indicate the picture corresponding to that sentence. The 76 sentences are classified according to the type of grammatical structure: locative, flexional, active affirmative, active negative, passive affirmative, passive negative, relative, and dative sentences. The score reflects the degree of impairment for each sentence, and a correct response is awarded 0 points. If a child fails the first attempt, the item is repeated and a score of 0.5 points is given for a correct response, whereas a second failure is given a score of 1 point. Thus, if a child fails at both first and second repetition his/her cumulative score will be 1.5 points (0.5 for the first error +1 for the second error). The total TCGB score ranged from 0 to 114.
Vocabulary development was assessed by means of an Italian version of the British Picture Vocabulary Scale (Dunn, Dunn, & Whetton, 1982). The scale is composed of 32 cards, each containing four pictures; the examiner enunciated a word, after which the child was asked to point to the picture that corresponded to the stimulus word. The score (0–32) represents the number of correct responses.
Visuomotor planning abilities were investigated by means of The Mazes subtest of the Wechsler Preschool and Primary Scales of Intelligence (WPPSI; Wechsler, 1973). The score represents exactly the raw score described in the WPPSI Manual.
Finally, attention and impulsive control were measured by means of two different tasks. The first was an auditory Continuous Performance Test, specifically devised for this study. The test includes 144 randomized stimuli; these comprise one target word “cane” (dog) and other nine words. The target word was presented in 36 stimuli, and the other randomized words accounted for the remaining 108 stimuli. There was a stimulus presentation every 2 s, and the child had to tap on the table when s/he detected the target word. An omission error was recorded if the child did not detect the target word (maximum number of omission was 36); a commission error was recorded if s/he tapped when a nontarget stimulus was presented (the maximum number of commissions was 108, that is, 144 stimuli minus 36 targets).
The second task was a Matching Familiar Figure Test (adapted from Marzocchi & Cornoldi, 1998): the child was presented with 12 cards in which the target picture was positioned at the top of the card and four other pictures were presented below. Only one of the four options was identical to the target (the other three contained minor differences). The child was asked to point to the picture identical to the target. If unsuccessful at the first attempt, s/he was invited to continue until the correct picture was found. The maximum number of errors was 36 (three possible wrong choices for 12 cards).
In addition, the parents were asked to use a rating scale (based on the attention-deficit/hyperactivity disorder symptoms reported in DSM-IV, American Psychiatric Association, 1994) to assess the degree of inattentive, hyperactive, and impulsive behaviors in their children. The scale included 14 items, and parents rated their children using a 4-point Likert scale from 0 (behavior not present) to 3 (behavior very frequent); the score ranged from 0 to 42 points.
The children's parents were invited to participate in a follow-up session. A home visit was arranged with the families who agreed to participate, and the order in which the tests were to be presented was fixed. During the visit the examiner carried out the complete battery of tests described above in a single session, which lasted approximately 1 hr.
RESULTS
Multiple comparisons between the two groups were carried out using the Mann–Whitney nonparametric test. As multiple significance tests are not usually independent, the Bonferroni method has been applied. We also calculated effect size for nonparametric statistical analyses (Kline, 2004). We have primarily based our comments on the results on the effect size as suggested by Kramer and Rosenthal (1999), who stated that, even when the results do not reach the conventional level of significance, by considering the effect size it is possible “to uncover a potentially interesting and valuable relationship that might have yielded more a significant result if only more subjects had been added to the study” (p. 64).
As can be seen in Table 2, the comparison between the two groups on grammar comprehension (TCGB) revealed that for five out of the eight grammatical structures and for the total score the effect size was large (>.80) even though the conventional level of significance was not reached. The LTs and the comparison groups differed significantly only on comprehension of the most difficult sentence type, that is, the passive negative (effect size = 3.18, which is very large), with the LTs obtaining significantly higher error scores than the ATs.
Table 2 also illustrates the receptive vocabulary development analysis. In this task the performance of the two groups did not differ significantly, although the effect size for this variable was large (.83).
The statistical analyses of visuomotor abilities, attention, and impulsiveness are reported in Table 3. The effect sizes were large in the Mazes Test (1.00), which evaluates visuomotor planning, and in the Matching Familiar Figure Test (1.37), which evaluates attention and impulsiveness.
A significant difference was also found between LTs and ATs on the nonword repetition task (effect size = 0.89), with the LTs obtaining significantly lower scores. The performance of the two groups did not differ on the Word Span Task, and the effect size was in the low range (see Table 4).
Phonological awareness analyses did not provide conclusive evidence because the 4-year-olds did not take part in the tasks and some of the older children were not able to complete them. The blending syllables task was completed by all 5- and 6-year-olds, and the mean scores obtained by the two groups were similar (LTs = 18.12, SD = 2.35; ATs = 19; SD = 1.93). Only four LTs and six ATs were able to complete the blending phonemes task, on which LTs obtained a mean score of 12.25 (SD = 8.50) and comparison children a mean score of 10.11 (SD = 11.06). Finally, six LTs and five ATs attempted the word segmentation tasks. The descriptive data showed a greater mean number of words segmented by ATs (M = 47.40, SD = 11.32) in comparison to LTs (M = 31.16, SD = 20.75).
DISCUSSION
The study reported here is one of the first attempts to investigate the development of LTs acquiring Italian, in a follow-up perspective and using a variety of neuropsychological measures, ranging from grammatical comprehension, vocabulary development, verbal STM, phonological awareness, planning and visuomotor coordination, and attention and impulsiveness.
The results of this study show that the linguistic competence of preschool Italian children, who were LTs at 24 months of age, was slightly different from that of their AT peers. These children achieved a lower score on the vocabulary development scale, and showed mild syntactic difficulties that are most marked in the comprehension of passive negative sentences, the most difficult grammatical form. This result is consistent with the findings reported in the literature showing that English-speaking LTs continue to perform at a lower level than their peers in a range of language-related abilities at a later age (Rescorla, 2002).
Even if statistical differences between the two groups did not always reach sufficiently large p values to use inferential statistics, the consideration of effect sizes revealed a pattern of results that encourages further studies with larger samples.
The evaluation of attention, impulsiveness, and visuomotor planning showed that the LTs did not present serious deficits in these neurocognitive domains. On an auditory continuous performance task in particular, the LTs did not significantly differ from controls on either attention (omissions) or impulsiveness (commissions). Furthermore, according to their parents, the LTs in this sample did not show significant behavioral problems such as inattention, hyperactivity, or impulsiveness. However, the performance of the LT group was lower than the control group on the Matching Familiar Figure and Mazes tests.
Considering that there are no studies in the literature that focus specifically on the LT's neurocognitive abilities investigated in this study, our results provide useful suggestions for future investigation.
The evaluation of verbal STM produced contrasting results, depending on the task considered. LT performance on the nonword repetition task was significantly lower than that of the control group, but they showed a similar level of competence to that of controls on the Word Span Task. This difference could be explained by hypothesizing that the nonword repetition task does indeed include a component of phonological awareness (Metsala, 1999), which is absent in the Word Span Task. However, the tasks that we devised to evaluate phonological awareness did not offer conclusive evidence of differences between the two groups because not all children completed them, and a large difference in scores was obtained only for the Word Segmentation task.
In conclusion, this study offers evidence that Italian LTs may be weak in several cognitive and linguistic areas that could constitute a significant risk factor for their academic achievements. Previous studies have suggested that slow early language development could lead to poor performance in a wide range of skills into middle childhood and into adolescence (Rescorla, 2000, 2002). Rescorla (2002) reports that children with a history of slow expressive language development obtained scores within the average range on most language tasks by age 5, but their performances on the language measures were significantly poorer through to age 9. Moreover, at 8 and 9 years children who were LTs had lower scores than comparison children on the aggregate reading measures (decoding, comprehension, spelling, and written language).
A more complete picture of the development of Italian speaking LTs could emerge once this research develops into a further follow-up study, focused on the children's reading, writing, and mathematical skills, when they reach 7 to 8 years of age.
APPENDIX A
APPENDIX B
